Proc. Natl. Acad. Sci. USAVol. 93, pp. 3984-3989, April 1996Neurobiology

The ALIAmide palmitoylethanolamide and cannabinoids, but notanandamide, are protective in a delayed postglutamate paradigmof excitotoxic death in cerebellar granule neurons

(N-methyl-D-aspartate/neurotoxicity/N-acylethanolamides/neuroprotection/receptor)

S. D. SKAPER*, A. BURIANI, R. DAL Toso, L. PETRELLI, S. ROMANELLO, L. FACCI, AND A. LEON

Researchlife S.c.p.A., Centro di Ricerca Biomedica-Ospedale Civile, 31033 Castelfranco Veneto (TV), Italy

Communicated by Rita Levi-Montalcini, Consiglio Nazionale Richerche, Rome, Italy, January 2, 1996 (received for review December 16, 1995)

ABSTRACT The amino acid L-glutamate is a neurotrans-mitter that mediates fast neuronal excitation in a majority ofsynapses in the central nervous system. Glutamate stimulatesboth N-methyl-D-aspartate (NMDA) and non-NMDA recep-tors. While activation ofNMDA receptors has been implicatedin a variety of neurophysiologic processes, excessive NMDAreceptor stimulation (excitotoxicity) is thought to be primarilyresponsible for neuronal injury in a wide variety of acuteneurological disorders including hypoxia-ischemia, seizures,and trauma. Very little is known about endogenous moleculesand mechanisms capable of modulating excitotoxic neuronaldeath. Saturated N-acylethanolamides like palmitoylethanol-amide accumulate in ischemic tissues and are synthesized byneurons upon excitatory amino acid receptor activation. Herewe report that palmitoylethanolamide, but not the cognateN-acylamide anandamide (the ethanolamide of arachidonicacid), protects cultured mouse cerebellar granule cells againstglutamate toxicity in a delayed postagonist paradigm. Palmi-toylethanolamide reduced this injury in a concentration-dependent manner and was maximally effective when added15-min postglutamate. Cannabinoids, which like palmi-toylethanolamide are functionally active at the peripheralcannabinoid receptor CB2 on mast cells, also preventedneuron loss in this delayed postglutamate model. Further-more, the neuroprotective effects of palmitoylethanolamide,as well as that of the active cannabinoids, were efficientlyantagonized by the candidate central cannabinoid receptor(CB1) agonist anandamide. Analogous pharmacological be-haviors have been observed for palmitoylethanolamide (ALI-Amides) in downmodulating mast cell activation. Cerebellargranule cells expressed mRNA for CB1 and CB2 by in situhybridization, while two cannabinoid binding sites were de-tected in cerebellar membranes. The results suggest that (i)non-CB1 cannabinoid receptors control, upon agonist binding,the downstream consequences of an excitotoxic stimulus; (ii)palmitoylethanolamide, unlike anandamide, behaves as an en-dogenous agonist for CB2-like receptors on granule cells; and(iii) activation of such receptors may serve to downmodulatedeleterious cellular processes following pathological events ornoxious stimuli in both the nervous and immune systems.

Dicarboxylic amino acids form the most widespread excitatorytransmitter network in the mammalian brain (1). Glutamateinteractions with specific membrane receptors are responsiblefor many neurologic functions, including cognition, memory,movement, and sensation (2). In addition, the excitationproduced by glutamate is important in influencing the devel-opmental plasticity of synaptic connections in the nervoussystem (3, 4). Excitatory amino acids (EAAs) have also beenimplicated in neurotoxicity. Excessive activation of EAA re-

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

ceptors may be responsible for much of the neuronal damageassociated with certain acute insults, including hypoxia-ischemia, hypoglycemia, epilepsy, and trauma (5, 6). Further-more, exaggerated EAA receptor activity has been suggestedby some to also underly chronic neurodegenerative disorders,including Huntington disease, Alzheimer disease, Parkinsondisease, amyotrophic lateral sclerosis, and acquired immunedeficiency syndrome-dementia complex (7-9). In many ex-perimental systems, overstimulation of the N-methyl-D-aspartate (NMDA) subtype of glutamate receptor with pro-tracted entry of Ca2+ into neurons appears to be a principalmechanism for subsequent damage (5). There is growingevidence for participation of non-NMDA receptors in EAA-mediated neurotoxicity as well, especially in cases of prolongedor chronic insult (10, 11).N-acylethanolamides, like palmitoylethanolamide, and N-

acylphosphatidylethanolamides accumulate in conditions in-volving degenerative changes to tissues (12), including brain(13) and cardiac (14) ischemia. Furthermore, EAAs canstimulate the synthesis of N-acylethanolamides and N-acylphosphatidylethanolamides in cultured central neurons(15, 16). The possibility that compounds of this type coulddefend against an excitotoxic insult may be entertained (16).Interestingly, it has recently been reported that mast cells,multifunctional immune cells implicated in immediate hyper-sensitivity and inflammatory reactions (17), express a periph-eral-type cannabinoid receptor (designated CB2; ref. 18) thatrecognizes palmitoylethanolamide (ALIAmides) and down-modulates activation of these cells in vitro (19). The candidateendogenous agonist for the brain cannabinoid receptor CB1(20, 21), arachidonylethanolamide (anandamide) (22, 23),binds to mast cell CB2 and actually antagonizes the functionaleffects of palmitoylethanolamide and several cannabinoids (19).We now report that palmitoylethanolamide and 2-O-(83-D-

glucopyranosyl)-N-palmitoylethanolamide, as well as somenatural and synthetic cannabinoids but not anandamide, areefficacious in protecting cultured cerebellar granule cells fromglutamate toxicity in a delayed postagonist paradigm withoutaffecting EAA receptor function. Anandamide, however, an-tagonized these neuroprotective effects. Furthermore, granulecells expressed mRNA for both CB1 and CB2 and cerebellarmembranes displayed two cannabinoid binding sites.

MATERIALS AND METHODS

Primary Neuron Cultures. Cultures containing granuleneurons were prepared from dissociated cerebella of 7- to8-day-old BALB/c mice (Modelli Biologici Sperimentali, Tre-viso, Italy) (24). Cells were plated in Eagle's basal medium

Abbreviations: EAA, excitatory amino acid; NMDA, N-methyl-D-aspartate; DIV, days in vitro; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; PMSF, phenylmethylsulfonyl fluoride;KA, kainic acid.*To whom reprint requests should be addressed.

3984

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996) 3985

supplemented with 10% fetal calf serum (BIOSPA, Wedel,Germany), 25 mM KCl, 2 mM glutamine, and gentamicin (50tag/ml) on 35-mm-diameter dishes (Falcon) coated with poly-(L-lysine) (10 Atg/ml) (Mr, 68,000), 2.5 x 106 cells per dish.Cytosine P3-D-arabinofuranoside (10 JLM) was added to theculture medium 18-20 h after plating to halt nonneurongrowth. The cultures were used at 8-10 days in vitro (DIV), andcontain -95% glutamatergic granule neurons (25).

Induction of Glutamate Neurotoxicity. Neurotoxicity wasinduced essentially as described (24). Medium from 8- to10-day-old granule cell cultures was removed and saved.Culture dishes were washed twice with Mg2+-free Locke'ssolution, and the cells were then incubated with 500 ,iMglutamate in Mg2+-free Locke's solution for 5 min (23-25°C).The glutamate-containing solution was then removed by as-piration, and the dishes were washed twice with completeLocke's solution and then returned to the incubator in theiroriginal medium for a further 24 h. Acute glutamate exposureunder these conditions consistently resulted in a 50-65% lossin neuron numbers 24 h later, at which time cell survival wasroutinely evaluated. Drug treatment protocols are described inthe appropriate figure or table legend as introduced. Stocksolutions of cannabinoids and N-acylethanolamides were madein dimethyl sulfoxide, while anandamide was dissolved inethanol. The final concentration of solvent in the culture neverexceeded 0.2%, except in the case of palmitoylethanolamide,which necessitated a higher concentration (1%). These con-centrations of solvent were found to have no effect on theresponse of the granule cells to glutamate.

Quantitation of Neurotoxicity. Glutamate neurotoxicity wasgross enough to be evident morphologically when viewedunder a phase-contrast microscope. Neuronal survival wasquantified using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltet-razolium bromide (MTT), which yields a blue formazan prod-uct in living cells but not in dead cells or their lytic debris (26).The reaction product, solubilized in dimethyl sulfoxide, iseasily measured with an ELISA plate reader and is directlyproportional to the number of neurons present (24, 27, 28).The MTT technique is equivalent to lactate dehydrogenaserelease in the measurement of excitotoxin-mediated neuronaldeath in vitro (29).

In Situ Hybridization. BALB/c mouse (4-week-old) brainwas snap-frozen in 2-methylbutane, 12-,im coronal sectionscut with a Jung model CM 3000 cryostat and thaw-mounted onpolylysine-coated glass slides. All sections were then fixed in4% paraformaldehyde, dehydrated in a graded series of eth-anols, incubated 5 min in chloroform, and air-dried. Cerebellargranule cells growing on polylysine-coated cover glasses at 8DIV were fixed with 4% paraformaldehyde, washed twice withphosphate-buffered saline, and permeabilized in 70% ethanolfor 48 h at 4°C. The cells were then dehydrated using a highergraded series of ethanols and air-dried. Detection of CB1- andCB2-specific mRNA made use of the following oligonucleo-tides prepared with a Beckman model Oligo 1000 DNAsynthesizer: CB1, 5'-GGT GAC GAT CCT CTT ATA GGCCAG AGG CCT TGT AAT GGA TAT GTA-3' (21); CB2,5'-GGT GAC GAG AGC TTT GTA GGT AGG TGG GTAGCA CAG ACA TAG GTA-3' (18). A random sequence wasused as control. Oligonucleotides were 3' end-labeled withdATP[a-35S] (1136 Ci/mmol; 1 Ci = 37 GBq; New EnglandNuclear) by terminal deoxynucleotidyltransferase (Pharma-cia) to a specific activity of 109 cpm/,tg. All cover glasseswere hybridized in standard solutions (30) with 1.5 X 107dpm/ml overnight in a humidified chamber at 42°C. The coverglasses were then washed once with lx SSC/0.1% SDS (30min, 55°C), twice with lx SSC (15 min, 55°C), and once with0.1x SSC (30 min, 25°C), followed by a 2-min rinse inautoclaved water and dehydration with a graded series ofethanols. The air-dried cover glasses were then dipped in K.5photoemulsion (Ilford) (diluted 1:1 with water), exposed for 5

weeks at 4°C, developed with Phenisol (Ilford), fixed withHypam (Ilford), and counterstained with cresyl violet.

Radioligand Binding Assays. Cerebella from 20-day-oldBALB/c mice were removed, cleaned of meninges, and storedat -80°C for up to 1 month. Membrane preparation andbinding assays were carried out following published proce-dures (19, 31), with modification. Groups of five (frozen)cerebella were homogenized in 20 ml of binding buffer (3 mMMgCl2/1 mM EDTA/50 mM Tris, pH 7.4) containing 0.32 Msucrose. The homogenate was centrifuged at 2000 x g for 10min at 4°C, and the resulting supernatant was centrifuged at15,000 x g for 15 min at 4°C. The pellet (P2) was resuspendedin 20 ml of binding buffer containing 0.1% fatty acid-freebovine serum albumin, and the last centrifugation step wasrepeated. The final pellet was gently rinsed with distilled waterand then resuspended in 1 ml of binding buffer (0.8-1.1 mg perml of protein). Membranes were kept on ice and used within1 h. Binding experiments were performed in silicon-treatedtubes. [3H]WIN 55,212-2 (45.5 Ci/mmol; 22 ttM in ethanol;New England Nuclear) and nonradioactive ligands were seri-ally diluted in dimethyl sulfoxide and were added at the desiredconcentration to a final volume of 500 ,ul of binding buffer. Thefinal concentration of dimethyl sulfoxide was always 1%.Binding was initiated by adding 30 ,ug (protein) of membranesand the tubes were incubated for 55 min at 30°C. Binding wasterminated by transferring the reaction mixture to an Eppen-dorf tube, followed by addition of fatty acid-free bovine serumalbumin to 0.1% and centrifugation at 39,000 x g for 10 min at20°C. The supernatants were collected and counted to determinethe concentration of free ligand, while the pellets were suspendedin 1% Triton X-100/ethanol (1:1; vol/vol) and radioactivity wasassayed by liquid scintillation counting. Nonradioactive WIN55,212-2 (1 ,tM) was used to displace specific binding.

Materials. Tissue culture media and supplements and can-nabidiol were obtained from Sigma. WIN 55,212-2 and A8-THC were from Research Biochemicals (Natick, MA). Palmi-toylethanolamide, 2-0-(f3-D-glucopyranosyl)-N-palmitoyleth-anolamide (glucosylpalmitoylethanolamide), and anandamidewere synthesized by standard chemical techniques, with purityof >99.5% as assessed by HPLC. All other reagents, unlessspecified otherwise, were from Sigma.

RESULTS

Palmitoylethanolamide- and Cannabinoid-Mediated Pro-tection from Glutamate Neurotoxicity. Although glutamate isa neuroexcitant agonist at NMDA, quisqualate, and kainatepostsynaptic receptors (32), its neurotoxicity is predominantlymediated by NMDA receptors alone (33). Given that palmi-toylethanolamide accumulates in ischemic brain (13) and itsproduction is induced in central nervous system neurons byEAA receptor activation (15, 16), it was of interest to examinepossible modulatory effects of the N-acylethanolamide againstNMDA receptor-mediated neurotoxicity. Most studies on theneuroprotective efficacy of glutamate antagonists or otherdrugs have used pre- and cotreatment protocols. The presentexperimental setting was chosen, as others have done (34, 35),to allow for conditions that provide potentially rescuableneurons in a delayed postagonist paradigm.

Palmitoylethanolamide (100 ,tM) or MK-801 (10 ,tM) wasadded during or at various times after conclusion of a 5-minglutamate exposure. Neuronal cell loss was quantitativelymonitored by the MTT reaction, which is widely used to assessneuron numbers in culture (24, 27, 28) and is consideredequivalent to lactate dehydrogenase release (29). Microscopicobservation of the cultures confirmed the results obtained byMTT (data not shown). The potent noncompetitive NMDAantagonist MK-801 (36) while maximally efficacious whenpresent during glutamate exposure rapidly decayed with a15-min postagonist delay (Fig. 1) (see also ref. 34). In contrast,

Neurobiology: Skaper et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996)

palmitoylethanolamide achieved maximal efficacy when added15 min postglutamate (Fig. 1). Neuroprotection produced bythe delayed, postagonist addition of palmitoylethanolamidewas also concentration dependent (Table 1) and related to theduration of drug exposure, being maximal when present for 20min or more. Addition of a glucose residue to palmitoyleth-anolamide (glucosylpalmitoylethanolamide) increased its ac-tivity as a neuroprotectant (Table 1), perhaps due to improvedsolubility. The potent non-NMDA receptor antagonist CNQX(11), Ca2+ channel blockers (diltiazem, nifedipine, o-cono-toxin, diphenylhydantoin, flunarizine), free radical scavengers,and monosialogangliosides were also tested in this delayedpostagonist rescue paradigm and found to produce little or noreduction in neuronal injury (data not shown).Given that cannabimimetics display functional effects anal-

ogous to palmitoylethanolamide on mast cells (19), cannabi-noid compounds were also explored for possible protectionagainst glutamate neurotoxicity. The synthetic cannabinoidnabilone (10 tLM) and A8-THC (10 taM), like saturated N-acyl-ethanolamides, were maximally effective with a delayed, 15-min postagonist application (Fig. 1); exposure times of 5-10min proved to be optimally efficacious. The respective EC50values are given in Table 1. In contrast, the nonpsychoactivecannabinoid cannabidiol, which has weak affinity for CB1 andCB2 receptors (18, 20), was inactive (Table 1).Anandamide Antagonizes the Neuroprotective Effects of

Saturated N-Acylethanolamides and Cannabinoids. Ananda-mide, a candidate agonist for the brain cannabinoid receptor(22, 23), produces many of the behavioral and physiologicalresponses of cannabinoids attributed to activation of the CB1receptor (37, 38). Unlike the active cannabinoids tested (Table1), a 10-min anandamide exposure of up to 100 ,tM (in thepresence of 150 ,M phenylmethylsulfonyl fluoride; PMSF)failed to protect granule cells from glutamate toxicity. Whenadded together with saturated N-acylethanolamides or canna-binoids, anandamide (10 ,tM) actually antagonized their de-layed postagonist neuroprotective effects, shifting the concen-tration-response curve for the drug (Table 1). The anandam-

120-

100 -

80-

60

40

20

0-CTRL - 5 0 1'5 30

MIN AFTER GLUTAMATE EXPOSUREFIG. 1. Time course of rescue from glutamate neurotoxicity:

palmitoylethanolamide and cannabinoids vs. MK-801. Sister cultureswere exposed to 500 ,iM glutamate for 5 min and then rescued byaddition of 10 ,iM MK-801 (0), 10 ,uM nabilone (A), 10 ,uM A8-THC(D), or 100 ,iM palmitoylethanolamide (A) at the indicated time inminutes after washout of the glutamate. 0, Glutamate alone. PMSF(150 ,uM) was included in all incubations. Except for those cultures inwhich the drug was present only during glutamate exposure (-5 on thetime axis), the duration of late addition drug treatment was as follows:MK-801, 30 min; palmitoylethanolamide, 30 min; cannabinoids, 10min. Neuronal survival by MTT was measured 24 h later. Values aremeans ± SD (three experiments).

Table 1. Palmitoylethanolamide and cannabinoidsconcentration-dependently rescue cerebellar granule cells fromglutamate toxicity with late addition: Antagonism by anandamide

EC50, ,LM

Compound Control + Anandamide

Palmitoylethanolamide 54.6 + 15.3 >150Glucosyl-PEA 12.1 ± 1.8 33.4 ± 3.5*Nabilone 3.9 ± 1.1 15.5 ± 0.6*A8-THC 2.8 ± 0.7 16.9 ± 5.1*11OH-A9-THC 0.88 ± 0.28 NDWIN 55,212-2 24.5 ± 4.1 NDCannabidiol > 100 NDAnandamide >100 ND

Cannabinoids were added to the cultures for a 10-min period andpalmitoylethanolamide or glucosylpalmitoylethanolamide (glucosyl-PEA) was added for a 30-min period, all starting 15 min postglutamate.All incubations contained 150 ,iM PMSF to inhibit anandamidemetabolism. Neuron survival was assessed 24 h later. EC5o is theconcentration reducing by 50% the cell death caused by glutamate.Values are means ± SD from at least three experiments. Anandamidewas used at 10 ,LM. ND, not determined.*P < 0.01 vs. control.

ide congener homo-,y-linolenylethanolamide (C20:3), whichbinds the CB1 receptor more weakly than anandamide (23),antagonized neither CB2 receptor function in mast cells northe rescue effects of saturated N-acylethanolamides and nab-ilone against glutamate neurotoxicity when used up to 25 juM(data not shown), indicating a degree of specificity in theobserved behaviors of anandamide.When granule cell incubation with anandamide (50 ,tM) was

prolonged beyond 1 h in the presence of 150 ,iM PMSF, aninhibitor of anandamide amidase (39), neuron survival wasreduced to 30% ± 12% (n = 6) of control 24 h later. Thepotent cannabimimetic compound HU-210 reportedly is cy-totoxic for neurons in vitro, an effect attributed to its interac-tion with the CB1 receptor (40). Active cannabinoids, but notcannabidiol, were also cytotoxic to granule cells when exposureto high (-25 piM) concentrations exceeded 60 min: 50 ,Mnabilone or A8-THC decreased survival by 80-85% after 24 h.Incubation of granule neurons with 100 ,uM palmitoylethanol-amide or its glucosyl derivative for 24 h did not affect cell vitality.

Palmitoylethanolamide and Cannabinoids Do Not Antago-nize EAA Receptor Function in Intact Neurons. The delayedpostglutamate neuroprotection afforded by saturated N-acylethanolamides and cannabinoids (Table 1) made unlikelya direct effect at the EAA receptor level, as confirmed here.The delayed component of kainic acid (KA)-mediated neu-ronal death involving free radical production especially evi-dent under conditions that eliminate the occurrence of acuteion-dependent neuronal swelling and cell lysis (40) was inves-tigated. When mature granule cells were treated with 500 utMKA for 30 min in a Na+-free buffer, the 75% loss seen 2 h laterwas, as expected (41), largely prevented by CNQX or theantioxidant butylated hydroxytoluene (Table 2). In contrast,neither palmitoylethanolamide nor cannabinoids were neuro-protective when present during and after KA exposure in theseconditions (Table 2). Palmitoylethanolamide and CNQX, butnot butylated hydroxytoluene, were cytoprotective for granulecells incubated with 500 ,tM KA in Na+-containing buffer, whererapid lysis of =50% of the neurons occurs during the 30-min drugexposure period (42). Thus, a KA receptor antagonist (CNQX)was able to reduce both components of KA receptor-triggeredneuronal injury while the N-acylethanolamide affected only theexcitotoxic branch, indicating that the latter operates indepen-dently of the EAA receptor.

Cerebellar granule cell development in vitro is strictly de-pendent on Ca2+ influx, mediated by either voltage- orNMDAreceptor-gated channels (43, 44). Predictably, MK-801 negated

3986 Neurobiology: Skaper et al.

v-

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996) 3987

Table 2. Palmitoylethanolamide and cannabinoids do not interferewith KA receptor activity in cerebellar granule cells in the absenceof Na+

Culture conditions % survival

KA 25.6 ± 3.1KA+CNQX 87.3± 11.1KA + BHT 84.6 ± 10.0KA + PEA 35.0 ± 4.2KA + nabilone 36.2 ± 4.9

Nine-day-old cultures were incubated for 30 min (27°C) in Na+-freeLocke's solution containing 500 ,uM KA and one of the following: 10,uM CNQX, 100 ,uM butylated hydroxytoluene (BHT), 100 ,uMpalmitoylethanolamide (PEA), 10 ,uM nabilone, or 3 ,uM 11OH-A9-THC. Cultures were then washed free of KA and incubated 2 h more(27°C) in Na+-free Locke's solution with the indicated compound.PMSF (150 ,uM) was present during all drug incubations. Cell survivalwas quantified at the end of the 2 h. Values are means ± SD (threeexperiments).

the trophic effect of NMDA for granule cells cultured 7 daysin medium containing a reduced concentration (5 mM) of KCI.In contrast, the presence of either glucosylpalmitoylethanol-amide or 11OH-A9-THC throughout the same 7-day period (inplace of MK-801) did not interfere with the NMDA trophicaction (Table 3). The latter cannabimimetic was chosen be-cause of its minimal long-term cytotoxicity. Palmitoylethanol-amide was not tested here, because of the deleterious influenceof long-term cell exposure to high solvent concentrations.

Cerebellar Granule Cells and Cerebellum Express theGenes Encoding Cannabinoid Receptors CB1 and CB2. Theobserved effects of cannabinoids and palmitoylethanolamidein downmodulating EAA neurotoxicity, and their antagonismby anandamide, prompted the question of whether cerebellargranule cells express cannabinoid receptor gene transcripts. Insitu hybridization of 8 DIV granule cell monolayers using CB1-and CB2-specific oligonucleotide probes revealed the presenceof mRNA for both receptor types in neuronal perikarya (Fig.2 A-C). The CB1 and CB2 mRNAs were detectable in themajority of granule cells examined, suggesting the same neu-ron to be capable of simultaneously expressing the two knowncannabinoid receptor subtypes. It was not possible, however, toascertain the relative abundance of the two mRNAs. Sectionsof mouse cerebellum, when hybridized with CB1- and CB2-specific probes, also revealed the presence of the correspond-ing mRNAs in the granule cell layer (Fig. 2 A'-C'). ThePurkinje cell layer, while producing a positive hybridizationsignal for CB2, was negative for CB1, indicating the ability ofthis in situ procedure to discriminate between different subsetsof neurons in the same tissue section.

Cerebellum Contains Two Cannabinoid Binding Sites. Ra-dioligand binding experiments demonstrated specific binding

Table 3. Glucosylpalmitoylethanolamide and cannabinoids do notinterfere with NMDA receptor function in developing cerebellargranule cells

Culture conditions % survival

K25 100KS plusNone 31.6 ± 3.8NMDA 69.6 ± 6.1NMDA + MK-801 33.3 ± 4.2NMDA + glucosyl-PEA 69.5 ± 1.5NMDA + 11OH-A9-THC 70.3 ± 2.9

Cultures were incubated from days 1-8 in medium containing 25mM KCl (K25) or 5 mM KCl (KS) and the indicated additions: NMDA,150 ,uM; MK-801, 10 ,uM; glucosylpalmitoylethanolamide (glucosyl-PEA), 50 ,tM; 11OH-A9-THC, 3 ,uM. The last two compounds wereresupplied daily. Cell survival was quantified at the end of the eighthday. Values are means ± SD (three experiments).

A * t._, _F.* v ___ _-

^ ^ s Xs_ v il_iPa_ ......

* j_SS. --,:- _!fi;_ ';iiliP r ._s,.s,.,

__-g!=.,:t,.. .:

o_ ... ... *._.

B4. ...

a .

s,

*.......:

4:* ::

ICI)

_1

FIG. 2. In situ hybridization of cannabinoid receptor mRNA incultured mouse cerebellar granule cells (A-C) and 4-week cerebellum(A'-C'). (A and A') CB1. (B and B') CB2. (C and C') Random. (Bar- 10 Jim.)

of [3H]WIN 55,212-2 to membranes from mouse cerebellum,with two apparent binding sites. The Scatchard (Rosenthal)plot is shown in Fig. 3. Values for Kdl of 1.6 ± 1.0 nM and Kd2of 11.0 ± 1.5 nM, with corresponding Bmx of 0.7 ± 0.1 and 3.2± 0.3 pmol per mg of protein, respectively, were obtained byanalyzing the data according to Rosenthal (45) and using theFIG. P computer program for a two binding site fitting (BIO-SOFT, Cambridge, U.K.). Prior studies of ligand binding tocannabinoid receptors in adult rat brain have described a single

0.06

0.05

0.04

, 0.03

0.02

0.01

aII

90-

75-

60-

lm45.-zo 300

Ca 15-

0 3000 6000 9000 12000 1S000 18000

FREE pM.

.

U

_-

_U

0 25 50 75 100 125 150 175 200

B

FIG. 3. Scatchard plot of specific [3H]WIN 55,212-2 binding tocerebellar membranes. Specific binding was defined as the differencebetween binding that occurred in the presence and absence of 1 JIMnonradioactive ligand. (Inset) Saturation isotherm. Data are means ±SD (seven experiments). B, bound (pM); F, free.

Neurobiology: Skaper et al.

i......

IOP.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1

3988 Neurobiology: Skaper et al.

binding site (31, 44). In these latter experiments, however,steps were taken (preincubation and washing) that were re-ported necessary for observing a homogeneous binding site(31), or that may have masked the presence of a higher affinitysite by the use of albumin (46). In initial trials anandamide fullydisplaced bound radioligand, while palmitoylethanolamideseemed to be only partially effective (data not shown). Giventhat WIN 55,212-2 has similar affinities for CB1 and CB2 (18,47), it is difficult to assign a site to palmitoylethanolamide,although it is reported to not bind CB1 (22, 23).

DISCUSSIONThe experiments described here assign a functional correlateto the naturally occurring saturated N-acylamide, palmi-toylethanolamide, in central neurons. We have demonstratedthat (i) palmitoylethanolamide and cannabinoids, but not theunsaturated N-acylamide anandamide, downmodulate in adelayed postagonist manner the toxic consequences of EAAreceptor activation in cultured cerebellar granule cells; (ii)anandamide antagonizes the neuroprotection afforded bypalmitoylethanolamide and cannabinoids; (iii) granule cellsexpress the genes for CB1 and CB2; and (iv) [3H]WIN 55,212-2binding to cerebellar membranes displays two different bind-ing affinities, suggesting the presence of at least two bindingsites. Because palmitoylethanolamide, but not anandamide,appears to behave as an endogenous agonist for the CB2receptor on mast cells and downregulates their activation invitro (19), the present findings suggest that a CB2-like receptormay also exert a negative regulatory effect on postglutamatereceptor events following excessive excitatory stimulation.

Palmitoylethanolamide, unlike anandamide, is reported tonot bind the CB1 receptor (22, 23). Anandamide, by bindingto brain CB1 (22, 23), produces many of the behavioral andphysiological responses of cannabinoids (47). The psychotropiceffects of cannabinoids are presumably mediated via activationof the brain CB1 receptor (31), which is a typical member ofthe G-protein-coupled superfamily of receptors (20, 21). Cul-tured granule cells and cerebellum expressed mRNA for bothCB1 and CB2 by in situ hybridization but other receptorvariants could exist. The binding data with cerebellar mem-branes are consistent with the presence of two binding siteswith affinities close to those published for CB1 and CB2 (18,47). Although CB2 receptor identification until now has beenlimited to peripheral tissues (18, 19, 48), a detailed study of itsexpression or that of other cannabinoid receptor forms indifferent brain areas as a function of species/development islacking. G-protein-linked cannabinoid receptors have beendescribed to be present in rat cerebellar granule cells in vitro(49), although receptor subtypes were not evaluated. Neuronalexpression of CB2 could be regulated by culture conditions, orin vivo by injury, as in the case of some EAA receptor subtypesin vitro in cerebellar granule cells (50) and after cerebralischemia (51). Palmitoylethanolamide, but not anandamide,was neuroprotective for granule cells; anandamide actuallyantagonized the protective action of the former. We recentlyreported that anandamide binds to the CB2 receptor on mastcells and antagonizes the ability of palmitoylethanolamide andcannabinoids to inhibit mediator release (19). Anandamidealso inhibited specific [3H]WIN 55,212-2 binding to mast cellmembranes (19), suggesting it to behave as a functionalantagonist at least for CB2 on mast cells. Such oppositebehavior is typical of differences in the agonistic ability ofreceptor ligands in cerebellar granule cells and, in analogy withmast cells (19), suggests that the two N-acylethanolamideshave differing roles for different cannabinoid receptors. In-terestingly, cannabinoid receptor activation may inhibit thepresynaptic release of glutamate via an inhibitory G protein inhippocampal neurons (52), suggesting a presynaptic CB1 lo-cation, thus further distinguishing the present delayed post-

agonist rescue effects from a CB1-mediated process. It is tempt-ing to speculate that saturated (ALIAmides) and unsaturated(anandamides) long-chain fatty acid ethanolamides may be can-nabinoid receptor type-specific endogenous agonists.The ability of glucosylpalmitoylethanolamide to prevent

excitotoxic neuronal injury in a delayed postagonist setting,and with greater efficacy than the parent molecule, may beattributable to one or several possibilities. Although modifiedsolubility may be a factor, metabolism or transformation of theglucosyl derivative to palmitoylethanolamide or a related N-acylamide needs to be considered. This avenue is currently beingexplored. While the glucosylacylamide exhibited a pharmacolog-ical profile not unlike that of the starting compound, it is presentlynot possible to draw conclusions as to whether or not the moleculeper se is recognized by cannabinoid receptors.

The observed neuroprotective effects of palmitoylethanol-amide and cannabinoids probably did not result from inter-ference with EAA receptor function. These compounds, incontrast to NMDA antagonists, were only modestly protectivewhen added concurrently with glutamate, but they becamemore efficacious with increasing delay of postagonist intro-duction. Saturated N-acylethanolamides and cannabinoids,however, reduced neither neuronal injury caused by kainatereceptor-induced oxidative stress nor neuronal survival pro-moted by NMDA receptor stimulation in immature cerebellargranule cells. As expected, the latter two processes weresensitive to kainate and NMDA receptor-specific antagonists,respectively. Moreover, palmitoylethanolamide and cannabi-noids prevented neurotoxicity triggered by NMDA or bykainate under conditions favoring excitotoxicity over freeradical toxicity. The nonpsychotropic cannabinoid HU-211attenuated NMDA receptor-mediated toxicity to culturedcortical neurons, apparently by binding to NMDA receptors(40). HU-211 was most effective when coapplied with EAAagonist (40). Antagonists of NMDA and kainate/AMPAreceptors, Ca2+ channel blockers, free radical scavengers,inhibitors of protein and RNA synthesis, phosphatases andnitric oxide synthase, and monosialogangliosides all failed toshow a delayed postagonist neuroprotective action (unpub-lished observations). The nonspecific endonuclease inhibitorATA attenuated glutamate neurotoxicity in cotreatment butnot when added 15-min postglutamate, in difference to arecent report of delayed ATA protection for cortical neurons(35). The N-acylamide and cannabinoid protective effects inthe delayed postagonist paradigm required only a limitedexposure window, suggesting these compounds to interferewith one or more downstream consequences of excitotoxicglutamate receptor overaction.

Given its clinical relevance, much effort has been directed todevising pharmacological means of preventing or reducing theneuropathological consequences of exaggerated EAA recep-tor stimulation, mainly by antagonizing receptor activation,blocking receptor- and voltage-gated ion channels, inhibitingEAA release, or by interfering with postreceptor processes(53-55). An alternative strategy to mitigate excitotoxic neu-ronal death would be to identify natural pathways whoseactivation downregulates the deleterious outcome of toxic ornoxious stimuli. In this respect, it is of more than passinginterest that EAAs themselves can induce brain neurons invitro to produce palmitoylethanolamide and other N-acylethanolamides (15, 16). Conceivably, certain of thesemolecules might play a role in modulating cellular defensemechanisms by acting at non-CB1 cannabinoid receptors,much as they do in the case of mast cells (19). By providing theneuron with exogenous palmitoylethanolamide (or semisyn-thetic saturated N-acylethanolamides), one might be makingavailable quantities of its physiological modulator sufficient torestore cellular homeostasis in the face of an excitotoxicchallenge. This pharmacologic approach has recently foundapplication in the ability of palmitoylethanolamide to control

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1

Proc. Natl. Acad. Sci. USA 93 (1996) 3989

mast cell activation by a local autacoid antiinflammatorymechanism, hence the term ALIA (19, 56, 57). Palmitoyleth-anolamide, in reality, appears to exert a more broad localautacoid antiinjury function, thus the acronym autacoid localinjury antagonism for ALIA. Therapeutic implications of thismechanism include the development of innovative neuropro-tective drugs for central nervous system injury.We thank Dr. Gabriele Marcolongo for synthesis and purification of

the N-acylethanolamides, Mr. Michele Fabris for preparation of thegraphs, Ms. Patrizia Lentola for secretarial assistance, and ModelliBiologici Sperimentali for skilled technical assistance. We extend aspecial note of thanks to the late Prof. Angelo Burlina, whose untiringefforts and initiative in biomedical research helped to make this workpossible.1. Cotman, C.W., Monaghan, D.T., Offersen, O.P. & Storm-

Mathisen, J. (1987) Trends Neurosci. 10, 273-280.2. Gasic, G. P. & Hollmann, M. (1992) Annu. Rev. Physiol. 54,

507-536.3. Collingridge, G. & Bliss, T. V. P. (1987) Trends Neurosci. 10,

288-294.4. Lipton, S. A. & Kater, S. B. (1989) Trends Neurosci. 12, 265-270.5. Choi, D. W. (1988) Neuron 1, 623-634.6. Choi, D. W. & Rothman, S. M. (1990) Annu. Rev. Neurosci. 13,

171-182.7. DiFiglia, M. (1990) Trends Neurosci. 13, 286-289.8. Meldrum, B. & Garthwaite, J. (1990) Trends Pharmacol. Sci. 11,

379-387.9. Lipton, S. A. & Rosenberg, P. A. (1994) N. Engl. J. Med. 330,

613-622.10. Koh, J. Y., Goldberg, M. P., Hartley, D. M. & Choi, D. W. (1990)

J. Neurosci. 10, 693-705.11. Sheardown, M. J., Nielsen, E. O., Hansen, A. J., Jacobsen, P. &

Hon6re, T. (1990) Science 247, 571-574.12. Schmid, H. H. O., Schmid, P. C. & Natarajan, V. (1990) Prog.

Lipid Res. 29, 1-43.13. Natarajan, V., Schmid, P. C. & Schmid, H. H. 0. (1986) Biochim.

Biophys. Acta 878, 32-41.14. Epps, D. E., Natarajan, V., Schmid, P. C. & Schmid, H. H. O.

(1980) Biochim. Biophys. Acta 618, 420-430.15. Di Marzo, V., Fontana, A., Cadas, H., Schinelli, S., Cimino, G.,

Schwartz, J.-C. & Piomelli, D. (1994) Nature (London) 372,686-691.

16. Hansen, H. A., Lauritzen, L., Strand, A. M., Moesgaard, B. &Frandsen, A. (1995) Biochim. Biophys. Acta 1258, 303-308.

17. Hollister, L. E. (1986) Pharmacol. Rev. 38, 1-20.18. Munro, S., Thomas, K.IL. & Abu-Shaar, M. (1993) Nature

(London) 365, 61-65.19. Facci, L., Dal Toso, R., Romanello, S., Buriani, A., Skaper, S. D.

& Leon, A. (1995) Proc. Natl. Acad. Sci. USA 92, 3376-3380.20. Matsuda, L. A., Lolait, S. J., Brownstein, M. J., Young, A. C. &

Bonner, T. I. (1990) Nature (London) 346, 561-564.21. Gerard, C. M., Mollereau, C., Vassart, G. & Parmentier, M.

(1991) Biochem. J. 279, 129-134.22. Devane, W. A., HanuS, L., Breuer, A., Pertwee, R. G., Stevenson,

L. A., Griffin, G., Gibson, D., Mandelbaum, A., Etinger, A. &Mechoulam, R. (1992) Science 258, 1946-1949.

23. Felder, C. C., Briley, E. M., Axelrod, J., Simpson, J. T., Mackie,K. & Devane, W.A. (1993) Proc. Natl. Acad. Sci. USA 90,7656-7660.

24. Skaper, S. D., Facci, L., Milani, D., Leon, A. & Toffano, G.(1990) in Methods in Neurosciences, ed. Conn, P. M. (Academic,San Diego), Vol. 2, pp. 17-33.

25. Levi, G., Aloisi, F., Ciotti, M. T. & Gallo, V. (1984) Brain Res.290, 77-86.

26. Mossmann, T. (1983) J. Immunol. Methods 65, 55-63.27. Manthorpe, M., Fagnani, R., Skaper, S. D. & Varon, S. (1986)

Dev. Brain Res. 25, 191-198.28. Ohsawa, F., Widmer, H. R., Knusel, B., Denton, T. L. & Hefti,

F. (1993) Neuroscience 57, 67-77.29. Patel, J., Zinkand, W. C., Thompson, C., Keith, R. & Salama, A.

(1990) J. Neurochem. 54, 849-854.30. Cosi, C., Spoerri, P. E., Comelli, C., Guidolin, D. & Skaper, S. D.

(1993) NeuroReport 4, 527-530.31. Devane, W. A., Dysarz, F. A., III, Johnson, M. R., Melvin, L. S.

& Howlett, A. C. (1988) Mol. Pharmacol. 34, 605-613.32. Watkins, J. C. & Olverman, H. J. (1987) Trends Neurosci. 10,

265-272.33. Choi, D. W., Koh, J.-Y. & Peters, S. (1988) J. Neurosci. 8,

185-196.34. Hartley, D. M. & Choi, D. W. (1989) J. Pharmacol. Exp. Ther.

250, 752-758.35. Csernansky, C. A., Canzoniero, L. M. T., Sensi, S. L., Yu, S. P. &

Choi, D. W. (1994) J. Neurosci. Res. 38, 101-108.36. Wong, E. H., Kemp, J. A., Priestley, T., Knight, A. R., Woodruff,

G. N. & Iversen, L. L. (1986) Proc. Natl. Acad. Sci. USA 83,7104-7108.

37. Crawley, J. N., Corwin, R. L., Robinson, J. K., Felder, C. C.,Devane, W. A. & Axelrod, J. (1993) Pharmacol. Biochem. Behav.46, 967-972.

38. Smith, P.B., Compton, D.R., Welch, S.P., Razdan, R.K.,Mechoulam, R. & Martin, B. R. (1994)J. Pharmacol. Exp. Ther.270, 219-227.

39. Deutsch, D. G. & Chin, S. (1993) Biochem. Pharmacol. 46,791-796.

40. Nadler, V., Mechoulam, R. & Sokolovsky, M. (1993) Neurosci.Lett. 162, 43-45.

41. Puttfarcken, P. S., Getz, R. L. & Coyle, J. T. (1993) Brain Res.624, 223-232.

42. Kato, K., Puttfarcken, P. S., Lyons, W. E. & Coyle, J. T. (1991) J.Pharmacol. Exp. Ther. 256, 402-411.

43. Gallo, V., Kingsbury, A., Balazs, R. & J0rgensen, O. S. (1987) J.Neurosci. 7, 2203-2213.

44. Balazs, R., Jorgensen, O. S. & Hack, N. (1988) Neuroscience 27,437-451.

45. Rosenthal, H. E. (1967) Anal. Biochem. 20, 525-532.46. Hillard, C. J., Edgemond, W. S. & Campbell, W. B. (1995) J.

Neurochem. 64, 677-683.47. Howlett, A. C. (1995) Annu. Rev. Pharmacol. Toxicol. 35, 607-

634.48. Lynn, A. B. & Herkenham, M. (1994) J. Pharmacol. Exp. Ther.

268, 1612-1623.49. Pacheco, M. A., Ward, S. J. & Childers, S. R. (1993) Brain Res.

603, 102-110.50. Aronica, E., Dell'Albani, P., Condorelli, D.F., Nicoletti, F.,

Hack, N. & Balazs, R. (1993) Mol Pharmacol. 44, 981-989.51. Paschen, W., Schmitt, J. & Uto, A. (1995) Soc. Neurosci. Abstr.

21, 765.52. Shen, M., Piser, T. M. & Thayer, S. A. (1995) Soc. Neurosci. Abstr.

21, 1569.53. Lipton, S. A. (1993) Trends Neurosci. 16, 527-532.54. Choi, D. W. (1990) Cerebrovasc. Brain Metab. Rev. 2, 105-147.55. Mercanti, D., Galli, C., Liguori, M., Ciotti, M. T., Gulla, P. &

Calissano, P. (1992) Eur. J. Neurosci. 4, 733-744.56. Aloe, L., Leon, A. & Levi-Montalcini, R. (1993) Agents Actions

39, C145-C147.57. Mazzari, S., Canella, R., Petrelli, L., Marcolongo, G. & Leon, A.

(1996) Eur. J. Pharmacol., in press.

Neurobiology: Skaper et aL.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 25

, 202

1